[Ffmpeg-devel] avformat/mpegts reads too much data for lowbitratestreams

[Måns Rullgård]

Michael Niedermayer <michaelni at gmx.at> writes:

> Hi
>
> On Thursday 09 June 2005 18:14, M?ns Rullg?rd wrote:
>> Kenneth Aafl?y <kenneth at aafloy.net> writes:
>> > torsdag 9. juni 2005, 17:27, skrev M?ns Rullg?rd:
>> >> Kenneth Aafl?y <kenneth at aafloy.net> writes:
>> >> > torsdag 9. juni 2005, 15:12, skrev M?ns Rullg?rd:
>> >> >> Kenneth Aafl?y <kenneth at aafloy.net> writes:
>> >> >> > How do you tell the user that an AVStream is not active anymore?
>> >> >>
>> >> >> You stop returning packets from it.  If all the streams go away, you
>> >> >> do whatever the method is to signal end of file, which I can't recall
>> >> >> at the moment.
>> >> >
>> >> > That's a crude interface, since there is no av_read_frame(AVStream)
>> >> > you have to depend on time to detect when to change what to decode!
>> >>
>> >> Huh?  If you get packets, you decode them.  If there are no packets,
>> >> you decode nothing.
>> >
>> > Say you have a ts with two services, and the user is currently
>> > decoding the first service. If this service stops it's transmission
>> > the user (app) either want to exit (since it's the end of the
>> > 'stream') or switch to the other service (since it's not the end of
>> > the 'file'). With your interface it's impossible (close to) to
>> > trigger either change, since there is no notion of 'end of
>> > substream'.
>>
>> Consider it a deficiency in libavformat.  Fixing it appears to be
>> nontrivial.
>
> hmm, whats non trivial from the lavf api point of view?
> you could add a flag to AVStream or AVPacket to indicate "end of stream"

Are you proposing a special end-of-stream packet?  I'm not against the
idea as such, but it would change the semantics.

-- 
M?ns Rullg?rd
mru at inprovide.com